Logistic regression
Dirk Speelman · Human cognitive processing · 2014
This text offers an introduction to binary logistic regression, a confirmatory technique for statistically modelling the effect of one or several predictors on a binary response variable. It is explained why logistic regression is exceptionally well suited for the comparison of near-synonyms in corpus data; the technique allows the researcher to identify the different factors that have an impact on the choice between near synonyms, and to tease apart their respective effects. Moreover, the technique is well suited to deal with the type of unbalanced data sets that are typical of Corpus Linguistics. First, we describe in which contexts logistic regression is applicable and we give examples of the types of research questions for which it is an appropriate tool. Next, we explain why and how logistic regression analysis is different from linear regression analysis and we illustrate how the output of logistic regression analysis can be interpreted, using the study of an alternation pattern in Dutch as our example. The R code used in the case study is explained in detail and an URL is given from which R code and data sets can be downloaded. Finally, suggestions for further reading are given.